Chinese DoorDash rival smashes open source record with 1.6-trillion-parameter LLM with a 1-million-context-token model crafted without Nvidia hardware
THE SO WHAT
A 1.6T-parameter, 1M-context model trained entirely on domestic chips shows that Nvidia is no longer a single point of failure for frontier-scale work. For operators, this is a concrete data point that geopolitical or export constraints won’t cap capability — they’ll just diversify the hardware and model ecosystem you have to integrate with.
READ THE SOURCE
MORE FROM THE WIRE
Applied AISpaceXAI Will Reportedly Release a Major New AI Model This Week
If SpaceXAI’s first joint model with Cursor is real and performant, you’re looking at an aerospace-grade data firehose meeting an agent-native IDE. For operators, that’s a hint that vertical labs plus agent tooling is a coming pattern — start mapping where your proprietary data plus an agent-first dev stack could create similar leverage.
Applied AIIf You Have a Public Instagram Account, You Might Be Surprised What AI Users Can Now Do With Your Face
Public social feeds are now de facto training and inference corpora for face-based AI — opting out is becoming an operational privacy decision, not just a UX preference. Brands and executives with high-visibility accounts should assume their likeness can be cloned and plan comms, legal, and verification accordingly.
Applied AIDeepSeek accidentally built a working ransomware strain, experts note, 'What we are witnessing is a fundamental shift in how novel cyber attacks are born'
A model casually generating working Android browser ransomware without human guidance means exploit ideation is now an AI-native workload. Security teams need to assume faster mutation cycles and broader attacker bench strength — and start testing their own defenses against model-generated attack patterns.
Applied AIMeta says Muse Image includes an invisible watermarking system called Content Seal and previews a web tool to check whether an image was generated with Meta AI
Invisible watermarking plus a public checker means provenance is becoming a product feature, not just a policy slide. If your org is generating synthetic media at scale, assume customers and regulators will expect this level of traceability within your own tools and workflows.